// engineering

Engagements.

Commercial work falls into four shapes. Every engagement starts with a scoping call, a written statement of work, and an NDA where appropriate. No tiered support, no offshored execution — you talk to the engineer doing the work.

[ AI-AUDIT ]

AI security assessment

Structured evaluation of an AI product's security posture — model, prompts, agent topology, tool boundaries, data handling. Output is a written report with severity-ranked findings and concrete remediation guidance.

↳ best fit: Companies shipping LLM products who need an outside read before launch, fundraising, or enterprise sales.

[ RED-TEAM ]

Adversarial red-team engagements

Time-boxed adversarial testing of a deployed AI system: jailbreaks, prompt injection, agent abuse, refusal-bypass, data exfiltration paths. Scoped to your threat model, not a generic checklist.

↳ best fit: Teams with an AI product in market or in late-stage build who want to know what an adversary would actually find.

[ INFRA ]

Secure AI infrastructure

Design and build of on-premise or air-gapped inference stacks for organizations that can't (or won't) send data to a hyperscaler. Covers model selection, hardware sizing, network isolation, and operational tooling.

↳ best fit: Healthcare, legal, defense-adjacent, and any team handling sensitive data who needs AI capability without the data-sharing tax.

[ ADVISORY ]

Technical advisory

Ongoing consultation for security or engineering leadership navigating AI risk — threat modeling, vendor evaluation, policy review, incident analysis. Retainer or per-engagement.

↳ best fit: CISOs, security leaders, and founders who want a knowledgeable second opinion in the room.

// process

How an engagement runs.

  1. 01

    Scoping call

    30–45 min, free. We discuss the system, the threat model, and what "good" looks like at the end.

  2. 02

    Statement of work

    Written scope, deliverables, timeline, and price. Fixed-fee or T&M. Mutual NDA.

  3. 03

    Execution

    Work happens on isolated lab hardware. Weekly progress notes. Critical findings escalated same-day.

  4. 04

    Report + walkthrough

    Written report (executive summary + technical detail), live walkthrough with your team, and follow-up retest of remediated findings included.

Start a conversation.

Reach out with a few sentences about what you're working on.

dev@aesirkode.com